Can an AI agent prepare management accounts?

Yes, once the month is properly closed. An AI agent can produce the profit and loss, balance sheet, cash flow, KPIs and variance tables, and draft findings, at depth 2. Two conditions decide it: the period's reconciliations are signed and the ledger is locked. The finance director reviews and approves the pack before anyone outside finance sees it.

Why does the close have to come first?

Management accounts are only as good as the ledger beneath them. An agent can produce a beautiful pack from a ledger with unreconciled suspense balances, missing accruals and miscoded costs, and the pack will be wrong with confidence. Speed makes this worse, not better: a pack produced the day after month end, before the close is finished, invites readers to act on numbers that will change.

So the agent should check before it starts. Are the balance sheet reconciliations signed? Are accruals posted? Is the period locked? If not, it reports what is outstanding instead of producing the pack.

What does the agent prepare?

• Statements. Profit and loss by department and in total, balance sheet, cash flow, all from the locked ledger and mapped through the management reporting structure built on the chart of accounts.
• Comparisons. Actual against budget, forecast and prior period, with variances above the agreed threshold highlighted.
• Drill-downs. For each material variance, the transactions driving it.
• KPIs. Calculated from defined formulas: margins, headcount costs, debtor and creditor days, using the same definitions every month.
• Working capital. Aged debtors and accounts payable summaries.
• Draft commentary. Findings, clearly separated from explanations, with questions for budget holders.

Where must the named approver sign off?

Before the pack leaves finance. The finance director, or a financial controller with that authority, reviews it and approves circulation. Their review covers three things the agent cannot settle: whether the numbers tell the right story, whether the explanations from budget holders hold up, and whether anything needs to be said that the numbers alone do not show.

Below that, each budget holder is responsible for the explanation of their own variances. The agent's draft goes to them first. The reasons are theirs, and the finance business partner challenges them. The full workflow for that step is in how to prepare variance commentary for management accounts.

An illustrative example

Take a chain of dental practices with 30 sites. On the third working day after month end, the controller locks the period. The agent checks that all reconciliations are signed and finds one outstanding: the patient deposits account at two sites. It waits, alerts the controller, and the reconciliation is signed that afternoon.

The agent then builds the pack. Gross margin is down at four sites. It traces the drop to higher laboratory costs and finds the same new laboratory supplier on all four. It drafts that finding and asks each practice manager to explain the switch. Three say the previous lab closed. The fourth gives no reason, and the business partner follows up.

The finance director reads the pack, adds a note on the laboratory supplier change across the group, and approves it for the leadership team.

What controls make it safe?

• No pack before close. The agent checks the close checklist first.
• Locked data. The pack is built from a snapshot of the locked ledger, so numbers do not shift after review.
• Fixed definitions. KPI formulas are stored and versioned.
• Explanations from owners. The agent drafts findings, not causes.
• Approval to circulate. Named, recorded, version-specific.

Where Pryme fits

In Pryme, producing the pack and sending it outside finance are separate capabilities. The agent can be certified to prepare the pack at depth 2; circulation is held for the finance director. Each version, check and approval is filed, so any number in a circulated pack can be traced back.

For groups, the pack follows consolidation of group accounts. The annual equivalent of this planning is covered in whether an AI agent can prepare a budget, and the wider controls in AI agents for finance: books you can audit.